Frequently asked questions

Getting started

What is Jantaris?

Jantaris gives you one account, one balance and one API key for dozens of AI models: image, video, music and audio, chat and code. You can use them in the browser with Jantaris Studio or call them from your own code through the Jantaris API. Both draw from the same prepaid balance, so you top up once and switch models whenever you like. The current list is on the models page.

Do I need to know how to code?

No. Studio runs in the browser: pick a model, write a prompt, set the duration or aspect ratio and click generate. The API is for developers and teams who want to plug models into their apps, scripts or automations. Both work on the same account and the same balance.

How do I get started?

Create an account, confirm your email with the link we send you, top up your balance (the minimum is $10) and either open Studio or create an API key in the dashboard. For the API, the quickstart takes you from a new key to your first generated image with ready-to-run curl, Python and JavaScript examples.

Is Jantaris free? Is there a free trial or free credits?

No. There is no free plan, no sign-up credits and no subscription - you pay from the first generation, and the smallest top-up is $10. In return your balance never expires and failed generations cost $0, so you don't pay for model errors. The pricing page shows the unit price of every model, which lets you work out in advance what a top-up covers.

Which models can I use?

Image models such as GPT Image 2.5, Nano Banana Pro and FLUX.2 [pro], video models such as Seedance 2.5, Kling 3.0 and Veo 3.1 (Quality), music with Mureka V9, and chat models such as Claude Opus 5.5, GPT-6 Sol, gemini-3.8-flash and DeepSeek V4.1 Flash. The catalog changes as new models come out. The up-to-date list with prices is on /models and available from GET /v1/models.

Does Jantaris work on my phone? Is there an app?

Yes, Studio and the dashboard work in a mobile browser such as Safari or Chrome. You can add the site to your home screen (Safari: Share → Add to Home Screen, Chrome: menu → Add to Home screen) and open it like an app. There is no native app in the App Store or Google Play yet. You can upload photos straight from your camera roll and save finished files with the Download button.

Can I use Jantaris without an account?

Generating needs an account, because every generation is paid from your balance. Without signing in you can browse the models page with parameters and samples, the pricing page and the docs, and fetch the catalog with prices from the public GET /v1/models endpoint without a key. Signing up takes a name, an email address and a password.

How is Jantaris different from the model makers' own apps, like Kling AI or Gemini?

Model makers run their own apps with monthly plans and usage limits, often with free trials too. Jantaris puts models from many makers in one place: one account, one balance, pay per image or per second of video with no subscription. You can try the same prompt in Kling 3.0, Veo 3.1 (Quality) and Seedance 2.5 without opening several accounts. Jantaris does not train its own models - each model page shows who made the model.

Studio

What can I create in Studio?

Images from text or from your own reference pictures, video clips from text or from a starting frame, music and audio, and conversations with chat models. Before each generation Studio shows what it will cost, and every finished task goes to your generation history. See the Studio page for an overview.

Can I start from my own photo or image?

Yes, if the model supports it. Many image models accept reference images for editing or keeping a consistent style, and many video models accept an image as the first frame of the clip. Each model page lists the inputs that model accepts. Only upload images you have the right to use, and see the rules on real people below.

How do I turn a photo into a video?

In Studio open the Video tab and pick a model that accepts an image as the first frame, such as Kling 3.0, Seedance 2.5 or MiniMax H3. Upload your photo with the "Upload file" button, describe the camera move and what should happen in the shot, then set a duration and an aspect ratio that matches the photo. You see the cost before you start. The image-to-video guide walks through it step by step, and how to make an AI video covers the whole path from idea to finished clip.

Can I write prompts in languages other than English?

Chat models understand and answer in many languages. Image and video models often handle other languages too, but English prompts usually give more predictable results, and some models document only English and Chinese. If a result drifts from what you described, translate the prompt to English and compare. The AI image prompt guide shows how to structure a prompt.

How long does a generation take, and can I close the tab?

Images are usually ready in under a minute and videos in a few minutes. The time depends on the model, clip length, resolution and current load. Tasks run on the server, so you can close the tab or lock your phone and find the result in your Studio history when you come back. Over the API you get the result by webhook or by polling GET /v1/tasks/{id}, see tasks.

Why did my generation fail?

Most often the prompt or image was rejected by the model maker's safety filter (for example because of public figures, brands, violence or nudity), the model does not support the parameters you chose, or there was a temporary outage. The task ends with status failed, Studio shows the reason marked "not charged", and the full hold returns to your balance. Rephrase the prompt, check the parameters on the model page or try another model, but do not try to get around the filters - our Terms of Service forbid it. API error codes are listed in errors.

Where are my files stored and for how long?

Your task history (prompt, settings, model and cost) stays in your account for 90 days. Output files are kept for 30 days after generation, so download anything you want to keep as soon as it is ready - Studio has a Download button for that. Jantaris is not a file storage service and an expired link cannot be restored.

Can I use generated images and videos commercially?

Jantaris claims no rights to your outputs. Whether you can use them commercially depends on the license of the specific model and on the law in the country where you use them: model developers publish their own terms, and in many countries purely AI-generated material may not be protected by copyright at all. You are responsible for what you generate and publish, including not infringing trademarks, protected characters or other people's rights. Before commercial use, check the model maker and its terms on the model page and read our Terms of Service.

Can I generate images or videos of real people?

Not without their consent. Deepfakes of real people made without consent, impersonation, intimate or sexual imagery of real people without consent, and any sexual content involving minors are prohibited and lead to account termination. Some models also refuse prompts that name public figures. Where the law requires it, for example under the EU AI Act transparency rules, label realistic AI-generated content as such when you publish it.

Choosing a model

Which model should I use for TikTok, Reels and Shorts?

For vertical 9:16 clips with sound, the usual picks are Seedance 2.5 (4-30 s, speech, sound effects and music), Kling 3.0 (3-15 s, lip-synced dialogue, several shots in one clip) and MiniMax H3 (4-15 s, stereo audio). Test the idea and the prompt cheaply at a lower resolution or in a lighter model such as Seedance 2.0 Mini, then render the final version in your target model. Camera moves and ready prompts are in AI video prompts and camera moves, and the differences between Seedance versions in Seedance 2.0 vs 2.5.

Which model is best for product photos and ads?

Start with the product shot: Nano Banana Pro takes up to 14 reference images and outputs up to 4K, while GPT Image 2.5 edits the shot from up to 16 reference images and closely follows long prompts. Then animate the finished shot as the first frame in a video model such as Kling 3.0 or Veo 3.1 (Quality). Always compare the logo, label and shape against the real product, because models can change them. The full process is in the AI product ad workflow, and the image models are compared in GPT Image 2.5 vs Nano Banana Pro vs FLUX.2.

Which model is best at text in images?

Text on posters, covers and packaging comes out well in GPT Image 2.5, FLUX.2 [pro] and Seedream 5.0 Pro, and the low-cost Z-Image Turbo ($0.02 per image) handles short English text. Put the exact words in quotes in your prompt and say where they go and in what kind of lettering. Accented letters and non-Latin scripts can come out wrong, so run a short test before a full batch. More in the AI image prompt guide and the Z-Image Turbo API guide.

Can a video have sound, dialogue or a voiceover?

Yes, several models generate audio together with the picture: Seedance 2.5 (speech, sound effects and music), Kling 3.0 (lip-synced dialogue and ambient sound), Veo 3.1 (Quality) (audio included in the clip price, no toggle) and MiniMax H3 (stereo audio described in the prompt). Write the lines a character should say word for word, in quotes. Speech quality varies between models and languages, so check it on a short clip before a longer project. Each model page shows whether the model supports audio.

Can I make a song or a music video?

Yes. Mureka V9 creates a full song from your lyrics (up to 3,000 characters) and a style prompt, returns the finished track as an audio file, and costs $0.124 per song. For a finished track, Music Video 1.0 makes a music video synced to the song - you upload an MP3 (10-300 s) and 1 to 7 character or style images. How to write lyrics and a style prompt is covered in the Mureka V9 music API guide.

Can I upscale or enhance a video?

Yes, in the Tools tab. Topaz Video Upscaler upscales a clip 2x and cleans it up, with no scale to choose (MP4, MOV or MKV up to 50 MB), priced by the length of the source clip: $0.11 per second. In Enhance Video 1.0 you pick a target resolution up to 8K (8K is not available on the lite tier) and a quality tier: lite, standard or professional; it also denoises and refines color. It helps when a video model gives you 720p and you need more.

API

Is the API compatible with the OpenAI and Anthropic SDKs?

Yes. Chat works through /v1/chat/completions in the OpenAI format (including stream: true) and through /v1/messages in the Anthropic Messages format. In most SDKs you only change the base URL to https://jantaris.com/v1 and pass your Jantaris key. Images, video and audio go through /v1/generate, described in the generate docs.

Can I use Jantaris with Claude Code, Cursor, n8n or Make?

Yes. Any tool that lets you set a custom base URL for the OpenAI or Anthropic API works with a Jantaris key. The Claude Code guide shows the environment variables to set, and the no-code guide covers n8n, Make and Zapier.

How much does Claude Code cost through Jantaris?

You pay per token with no subscription: Claude Sonnet 5 costs $3.24 / $16.20 per 1M tokens and Claude Opus 5.5 costs $5.40 / $27.00 per 1M tokens. Your monthly cost depends on how long your sessions are and how much code the agent reads - the dashboard shows your spend per model for the last 30 days. Create a separate key with a monthly spend limit for Claude Code. Setup is in the Claude Code docs, and the comparison with a subscription in Claude Code cost: pay per token.

How does video generation work over the API?

Video and audio tasks are asynchronous. POST /v1/generate returns a task with status queued and an estimated_cost, and you then poll GET /v1/tasks/{id} or receive the finished task on your webhook_url. For short jobs such as images you can add ?wait=60 (up to 120 seconds) and get the result in the same response. See tasks and webhooks.

What are the rate limits?

By default each API key can start 60 generations per minute. Above that the API returns 429, so wait and retry with exponential backoff instead of retrying in a tight loop. If your project needs a higher limit, contact us with your expected volume and use case. All error codes are listed in errors.

How do I set a spend limit on an API key?

When you create a key in the dashboard (API keys section), enter a monthly limit in your account currency - the field is optional. If a request would go over the limit, the API returns 403 with type permission_error and your account balance is left untouched. Give each automation, test environment and coding tool its own key with a limit, so a runaway script cannot spend your whole balance. Details in authentication.

How do I keep my API key safe?

Treat the key like a password: keep it on the server, in environment variables or a secrets manager, and never put it in front-end code, mobile apps or public repositories. Use separate keys for separate projects so you can revoke one without breaking the rest. If a key leaks, revoke it in the dashboard right away and create a new one, because you are responsible for usage made with your keys until they are revoked. For webhooks, verify the x-jantaris-signature header (HMAC-SHA256 of the body) before trusting the payload. More in authentication.

Can I build my own app or client service on the Jantaris API?

Yes. The API can be the back end of your app, store or automation, and you sell your own product to your customers. Keep the key on your server and never hand it to customers. Our Terms of Service do not allow reselling raw access to Jantaris or sharing keys in a way that lets others use the service without their own account, unless we have agreed to it in writing. Also check the terms of the model makers your service relies on.

What happens if a model provider has an outage?

For some models we can run the same model through a second infrastructure provider and switch automatically, so a single outage does not stop your requests. When no route is available, the API returns 502 and you are not charged for that task. Build your client to retry 502 and 429 with backoff.

How do I check my balance from code?

Call GET /v1/balance, which returns the balance in your account currency. If your balance is too low to cover the estimated_cost of a task, the API answers 402 before anything starts, so there are no half-finished charges. Billing details are in billing docs.

Models & pricing

How are prices calculated?

Every model has a public unit price: per image, per second of video, per track or second of audio, or per million input and output tokens for chat. For example Seedance 2.5 costs from $0.261 per second, GPT Image 2.5 costs $0.0432 per image and Claude Sonnet 5 costs $3.24 / $16.20 per 1M tokens. Before a generation starts you see its estimated_cost in Studio and in the API response, and that amount is held on your balance. Chat is billed on the tokens actually used. All prices are on the pricing page.

How much does an AI video or image cost on Jantaris?

Multiply the model's unit price by the number of seconds or images. Video costs from $0.169 per second in Kling 3.0, from $0.261 per second in Seedance 2.5 and from $0.161 per second in MiniMax H3, while Veo 3.1 (Quality) is billed per clip: from $2.53 per video. An image costs, for example, from $0.0747 per image in Nano Banana Pro or $0.02 per image in Z-Image Turbo. Prices per resolution and worked examples are in the guides to Kling 3.0 API pricing, Veo 3.1 API pricing, MiniMax H3 (Hailuo), Grok Imagine and Gemini Omni.

Which video model is the cheapest?

The lowest rate belongs to Veo 3.1 Lite: from $0.177 per video, which covers a whole 8-second clip. Among models billed per second, the cheapest are Grok Imagine 1.0 Video (from $0.0299 per second), Seedance 2.0 Mini (from $0.0366 per second) and PixVerse V6 (from $0.0405 per second), plus Seedance 2.0 Mini Eco (from $0.0451 per second) at 720p. They work well for testing ideas and prompts, and it is worth comparing motion, detail and sound with higher-end models such as Kling 3.0 or Veo 3.1 (Quality) afterwards. Many models also charge less at lower resolutions. Among image models, Z-Image Turbo and Grok Imagine are among the cheapest, and every rate is listed side by side on the pricing page. The AI cost calculator shows the cost of a 5 and 10 second clip in every model at once.

Why does Studio show a reserve for vocal removal or video upscaling?

These tools are billed by the length of the file you upload: Vocal Remover per minute of audio, Topaz Video Upscaler per second of video. Studio reads the file length and shows an estimate, while the balance holds a buffer (10 minutes of audio, 120 seconds of video, 10 seconds for Motion Control). When the job finishes we charge the actual amount and release the rest of the hold right away.

Is prompt caching cheaper on chat models?

On some models, yes. GPT-6 Sol, GPT-6 Luna, Claude Haiku 4.5 and Kimi K3 bill cached input at a separate, several times lower rate shown on the model page. On other models cache reads are billed at the input rate and cache writes at 1.25x the input rate.

Do I pay for failed generations?

No. When a task ends with status failed, its cost is 0 and the amount held when you started it is released in full. This includes provider errors, timeouts and prompts rejected by the model. A task that succeeds is charged even if you don't like the result, because the model did the work. After 25 failed generations within 24 hours we pause new tasks from the account (error 429) until older failures drop out of that window.

Is there a subscription or a monthly fee?

No. Jantaris works on a prepaid balance with no subscription and no minimum commitment. You pay for what you generate, and the balance does not expire.

Which currency do you charge in?

Accounts on jantaris.com are billed in US dollars, and prices there exclude taxes. The amount you pay is the amount added to your balance. The Polish version at jantaris.com/pl (the short address jantaris.pl redirects there) bills in Polish zloty with 23% VAT included, supports BLIK and issues Polish VAT invoices. The currency is set for your account when you sign up, and balances are not converted between the two versions.

Can model prices change?

Yes. A price can change when a provider changes its rates or a model gets a new version, and the current price is always on the model page and in GET /v1/models. A task you have already started is charged at the price shown when you started it.

Payments & invoices

How can I pay?

On jantaris.com you top up by card. Payments are processed by Stripe, so your full card number never reaches our servers, and your balance updates as soon as Stripe confirms the payment, usually within seconds. Customers in Poland can use the Polish version to pay in zloty with BLIK or a card.

What top-up amounts are available?

The minimum top-up is $10. Ready-made packages are $10, $25, $50, $100 and $250, or you can enter any amount from $10 up. Top-ups carry no bonuses - your balance grows by exactly what you pay. Compare the options on the pricing page.

Does my balance expire?

No. Your balance stays on your account until you use it, with no expiry date and no inactivity fee.

Will I get an invoice?

Yes. An invoice or receipt is issued for every top-up and is available in the billing section of your dashboard. If you buy as a business, add your company name, address and VAT ID in the billing details before you pay, because they are printed on the document at the moment it is issued.

Do you charge VAT or sales tax?

Prices on jantaris.com are in US dollars and exclude taxes. The amount you pay at checkout equals the amount added to your balance. If you need a Polish VAT invoice, or reverse charge for an EU business with a valid VAT number, use the Polish version, which bills in Polish zloty.

Can I get a refund?

Yes, within 14 days of a top-up. Write to us from the email address linked to your account: if you have not used the funds from that top-up, we refund the full amount, and if you have used part of them, we refund the unused part. The refund goes to the original payment method within 14 days of your request, and referral credits are not refunded. After 14 days top-ups are generally not refundable (the exceptions, such as us ending the service, are in our Terms of Service), but your balance never expires. Failed generations are never charged in the first place.

Privacy & security

Do you train AI models on my prompts or outputs?

No. The content of your prompts and outputs is not used to train models. To create a result, your prompt is sent to the infrastructure that runs the model you picked, and it is used only to process that request.

Who can see my prompts and how long do you keep them?

Prompts and settings are kept in your task history for 90 days so you can see past work and so we can detect abuse. Technical logs are kept for 30 days and billing records for 5 years, as tax law requires. We look at the content of a task only to handle a support request you sent us or to investigate a suspected violation of our rules. The full picture is in the Privacy Policy.

Can other users see my generations?

No. Jantaris has no public gallery - your task history and prompts are visible only in your account. A link to a file, whether an output or an upload, works for anyone you share it with while it is valid, so treat it as private. How we store data is described in the Privacy Policy.

What content is not allowed?

Illegal content of any kind, any sexual content involving minors, intimate imagery of real people without their consent, deepfakes of real people without consent, impersonation of people or organizations, malware, spam and phishing. Automated filters may block a request, and serious or repeated violations lead to account termination. The full list is in the acceptable use section of our Terms of Service.

How do I delete my account?

Write to us from the email address linked to the account (the contact address is in the Terms) and we will delete the account within 30 days. We delete your account data and task history. Billing records stay for the period required by tax law. Use your balance or request a refund before you delete, because a balance left on a deleted account cannot be recovered.

Referral program

How does the referral program work?

Copy your referral link from the dashboard (it ends with ?ref=YOURCODE) and share it. For every person who signs up through it, you get 20% of what they spend on image, video and audio models and 5% of what they spend on chat models, for 12 months from their sign-up. Commission counts from the referred person's first paid top-up and is added to your Jantaris balance. Details on the affiliates page.

How much can I earn from referrals?

For example, if someone you referred spends $250 on video and $50 on chat within 12 months of signing up, you get $50 + $2.50 = $52.50 on your balance. Commission is based on what they actually spend, so their failed generations ($0) do not count. You can spend it on any model in Studio or through the API.

How long does a referral link stay valid?

After someone clicks your link, a referral cookie remembers your code for 30 days. If they sign up within that time, they are counted as your referral. If they sign up later or after clearing cookies, the referral cannot be assigned.

Can I withdraw referral earnings as cash?

No. Referral commission is credited to your balance, and you can spend it on any model in Studio or through the API. It is not refundable and cannot be transferred to another account.

Can I refer myself or my own accounts?

No. Self-referrals, duplicate accounts and sign-ups made without the referred person's knowledge are not counted. If we find abuse, we remove the commission and may close the accounts involved.